{
  "id": 231974,
  "title": "One model - one building VS one model - all buildings",
  "url": "/competitions/indoor-location-navigation/discussion/231974",
  "author_name": "",
  "post_date": "2021-04-11T14:44:33.759082200Z",
  "votes": 4,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Most people build models based on wifi features to predict the coordinates (as a base for further postprocessing).  I am curious what works better for you: one model per each building or one model for all the buildings?</p>",
  "messages": [
    {
      "id": "1270338",
      "postDate": "04/11/2021 14:44:33",
      "content": "<p>Most people build models based on wifi features to predict the coordinates (as a base for further postprocessing).  I am curious what works better for you: one model per each building or one model for all the buildings?</p>",
      "rawMarkdown": "Most people build models based on wifi features to predict the coordinates (as a base for further postprocessing).  I am curious what works better for you: one model per each building or one model for all the buildings?",
      "votes": null
    },
    {
      "id": "1270683",
      "postDate": "04/11/2021 22:03:40",
      "content": "<p>One model for all, more data</p>",
      "rawMarkdown": "One model for all, more data",
      "votes": null
    },
    {
      "id": "1271216",
      "postDate": "04/12/2021 12:07:15",
      "content": "<p>I've done both ways and didn't notice a significant advantage either way - there are pros and cons to each approach. One model (all buildings) is much easier to handle though (than running or training a model per site), so that's a significant reason to use just one :)</p>",
      "rawMarkdown": "I've done both ways and didn't notice a significant advantage either way - there are pros and cons to each approach. One model (all buildings) is much easier to handle though (than running or training a model per site), so that's a significant reason to use just one :)",
      "votes": null
    },
    {
      "id": "1273491",
      "postDate": "04/14/2021 11:54:12",
      "content": "<p>I'm working on each building approach now. It's not efficient but might pay off in the end. </p>",
      "rawMarkdown": "I'm working on each building approach now. It's not efficient but might pay off in the end.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1270683,
      "author_name": "enric1296",
      "author_url": "",
      "post_date": "04/11/2021 22:03:40",
      "content": "<p>One model for all, more data</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1271216,
      "author_name": "chris62",
      "author_url": "",
      "post_date": "04/12/2021 12:07:15",
      "content": "<p>I've done both ways and didn't notice a significant advantage either way - there are pros and cons to each approach. One model (all buildings) is much easier to handle though (than running or training a model per site), so that's a significant reason to use just one :)</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1273491,
      "author_name": "crained",
      "author_url": "",
      "post_date": "04/14/2021 11:54:12",
      "content": "<p>I'm working on each building approach now. It's not efficient but might pay off in the end. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1270338": "Most people build models based on wifi features to predict the coordinates (as a base for further postprocessing).  I am curious what works better for you: one model per each building or one model for all the buildings?",
    "1270683": "One model for all, more data",
    "1271216": "I've done both ways and didn't notice a significant advantage either way - there are pros and cons to each approach. One model (all buildings) is much easier to handle though (than running or training a model per site), so that's a significant reason to use just one :)",
    "1273491": "I'm working on each building approach now. It's not efficient but might pay off in the end."
  },
  "source": "meta"
}